JP2006220487A - Apparatus and method of estimating position and position estimating program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To presume simply the position of a moving object having an oscillator in a field which arranges a plurality of receivers. <P>SOLUTION: In order to presume the position of the object having the oscillator in the field where the plurality of the receivers are arranged: from the sensing recording data which recorded the sensing circumstances of the above oscillator by each above receiver on time series, the group of the feature vector representing the sending circumstances of the above oscillator by each of the plurality of the receivers in unit of time is formed from the sensing write data which recorded the sensing circumstances of the above oscillator by each above receiver on time series acquired beforehand; the grouping of the group of the above feature vector is performed on the basis of the predetermined grouping nominal, and the feature vector which represents the group to each group is formed; and a label attachment is performed in the location in the above field for every representative feature vector. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a position estimation device, a position estimation method, and a position estimation program.

Currently, the development of a system that obtains the position and flow line of a moving body indoors from the transmitter that sends a weak radio wave given to the moving body and the position of the receiver placed in the field is underway. Techniques related to these are proposed in, for example, Patent Document 1. In the moving body position detection system, the moving body is provided with a transmitter, and a receiver is installed so as to cover the entire field centering on a place where it is desired to know that the moving body has visited. When the mobile object comes close to the receiver (observation spot), the receiver receives the radio wave of the transmitter of the mobile object, and specifies the position of the mobile object from the position of the receiver that has received the radio wave of the transmitter.
Japanese Patent Application No. 2002-378609

  In a moving body position detection system using a transmitter, especially when installing many observation spots in a narrow field and observing the movement trajectory of the moving body, a receiver that is in close proximity and the same transmitter that the moving body has simultaneously You may receive radio waves transmitted from. At this time, there arises a problem that the moving body having the transmitter does not know which observation spot is located near the center.

  This problem can be avoided by adjusting the sensitivity of the antenna of the receiver and the position of the observation spot to be installed so that only one receiver receives the radio waves from each transmitter. Takes a lot of work. In addition, since the shape of the observation spot is determined by the field environment and cannot be freely deformed, there is a problem in that the observation spot is made small and a large number of receivers are required to cover the entire field. In addition, when multiple receivers receive radio waves from the same transmitter, a method of determining each observation spot where the user is located according to the combination of receivers that simultaneously received radio waves can be considered. When there are many various combinations, it is difficult to define observation spots for all the combinations. Also, the transmitter's transmitted radio waves are received when the mobile object has multiple transmitters due to various factors such as the amount of battery installed in the transmitter and the characteristics of the transmitter, and the environment. In a machine, not all of the transmitters that are possessed are detected. In such a case, it is difficult to determine which observation spot the mobile body is in the vicinity of.

  The present invention has been made in order to solve the above-described problems, and a position estimation device and position estimation that can easily estimate the position of a moving body having a transmitter in a field in which a plurality of receivers are arranged. A method and a position estimation program are provided.

  A position estimation apparatus according to an aspect of the present invention is a position estimation apparatus that estimates a position of a mobile body having a transmitter in a field in which a plurality of receivers are arranged: each reception received in advance A feature vector generation unit that generates a group of feature vectors representing the detection status of the transmitter by each of the plurality of receivers in a unit time from detection record data in which the detection status of the transmitter by a device is recorded in time series A representative feature vector generation unit that groups the group of feature vectors based on a predetermined grouping criterion and generates a representative feature vector representing the group for each group; and for each representative feature vector A position estimation reference generation unit for labeling a place in the field.

  A position estimation method according to an aspect of the present invention is a position estimation method for estimating the position of a mobile body having a transmitter in a field in which a plurality of receivers are arranged: Generating a group of feature vectors representing the detection status of the transmitter by each of the plurality of receivers in a unit time from detection record data in which the detection status of the transmitter by a device is recorded in time series; Grouping based on a predetermined grouping criterion and generating a representative feature vector representing the group for each of the groups; labeling a location in the field for each representative feature vector; It is characterized by.

  The position estimation program as one aspect of the present invention is a position estimation program that causes a computer to function as means for estimating the position of a mobile object having a transmitter in a field in which a plurality of receivers are arranged: A group of feature vectors representing the detection status of the transmitter by each of the plurality of receivers in a unit time is generated from detection record data in which the detection status of the transmitter by each acquired receiver is recorded in time series Means for grouping the group of feature vectors based on a predetermined grouping criterion, and generating a representative feature vector representative of the group for each group; for each representative feature vector, in the field The computer as a means for labeling

  According to the present invention, it is possible to easily estimate the position of a moving body provided with a transmitter in a field where a plurality of receivers are arranged.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 is a block diagram schematically showing a configuration of a position estimation apparatus as an embodiment of the present invention.

  The position estimation apparatus includes a receiver acquisition data storage unit 1, a mobile-object-owned transmitter ID storage unit 2, a feature vector conversion unit 3, a position estimation reference generation condition setting unit 4, a position estimation reference generation unit 5, and a position estimation reference storage. Unit 6, position estimation unit 7, movement restriction data storage unit 8, position determination reference setting unit 9, estimated position determination unit 10, alternative generation condition setting unit 11, alternative generation unit 12, and action history storage unit 13 . The functions of all or part of these functional units may be realized by causing a computer to execute a program generated using a normal programming technique, or may be realized by hardware.

  The receiver acquisition data storage unit 1 records the radio wave observation status of the transmitter by the receiver and stores it as receiver acquisition data.

  FIG. 2 shows an example of the contents of the receiver acquisition data storage unit.

  The receiver acquisition data storage unit 1 is composed of a plurality of event records. The event receives the radio wave of the transmitter that has not been received until now, that is, when the transmitter that was not in the observation spot moves into the observation spot and the radio wave of the transmitter that has received the radio wave until now Occurs when the transmitter moves out of the observation spot, and is recorded as an event record.

  The event record is a radio wave status identification to identify the radio wave acquisition status (whether the transmitter that was not received so far has been received or whether the radio wave of the transmitter that has been received so far has been received) Code (AorL), transmitter ID (TAGID001, TAGID002, ...) for identifying the transmitter that is transmitting the radio wave, receiver ID (L1, L2, L3,...) And event occurrence time (Time1, Time2, Time3, Time4,...) Indicating the time when the event occurred.

  The event records are stored in the receiver acquisition data storage unit 1 in time series. For example, the first line of the receiver acquisition data indicates that the transmitter TAGID001 has visited the observation spot L1 at time Time1. The fourth line indicates that the transmitter TAGID001 has left the observation spot L1 at time Time4. That is, the first and fourth lines indicate that transmitter TAGID001 stayed at observation spot L1 from time Time1 to time Time4.

  In FIG. 1, one or more transmitter IDs of transmitters possessed by the mobile body are recorded in the mobile body possessed transmitter ID storage unit 2 for each mobile body.

  FIG. 3 shows an example of the contents of the mobile-owned transmitter ID storage unit 2.

  The mobile object possessed transmitter ID storage unit 2 is composed of a plurality of mobile object records. A mobile record is composed of a mobile ID and one or more transmitter IDs. For example, the first line indicates that the mobile unit PID001 possesses three transmitters TAG001, TAG002, and TAG003. In addition, it is good also considering what added the attribute information of a mobile body, such as sex, work content, and age, a questionnaire result, etc. to these contents as a mobile body record.

  In FIG. 1, the feature vector conversion unit 3 is based on the event records stored in the time series in the receiver acquisition data storage unit 1 and the mobile unit record stored in the mobile unit possessed transmitter ID storage unit 2. A feature vector record is generated in time series for each moving object, and the generated feature vector record is output to the position estimation reference generation unit 5 and the position estimation unit 7.

  FIG. 4 is a diagram illustrating an example of the content of a feature vector record.

  The feature vector record includes a moving body ID for identifying the moving body, a time, and a feature vector (radio wave reception status data) representing the reception status of the radio waves of the transmitter by the receiver.

  The number of elements of the mobile object ID and time is 1, and the number of elements of the radio wave reception status data is the total number of receivers installed in the field.

  The element of the radio wave reception status data stores the number of transmitters detected by the same moving body by each receiver in a predetermined time (unit time). The first line shows that the mobile unit PID001 receives radio waves from TL1001 transmitters at the receiver L1 and the radio waves from TL2001 transmitters at the receiver L2 at time Time001 (between time Time001 and the reference time). This indicates that the receiver L3 has received radio waves from TL3001 transmitters. (Time001-reference time), (Time002-Time001), and (Time003-Time002) are equal, and each corresponds to a unit time. Feature vector records are generated (total observation time / unit time) for one mobile object. When the moving body has a plurality of transmitters, the radio wave reception status data of the feature vector may be normalized (for example, the number of detections is divided by the number of possessions).

  In FIG. 1, the position estimation reference generation condition setting unit 4 sets parameters (reference generation parameters) necessary for generating a position reference record, which will be described later, based on user selection, for example.

  As the reference generation parameters to be selected by the user, a similarity parameter for calculating the similarity between vectors, a parameter for indicating a method for generating a transmitter group (described later), and conditions necessary for generating the transmitter group are determined. There are parameters for. The conditions necessary for generating the transmitter group differ depending on the group generation method and may be directly input by the user.

  The similarity between the vectors is an index representing how similar two vectors are. As the similarity, for example, there is a distance between two vectors. Depending on the distance calculation method, there are Euclidean distance, Euclidean square distance, standardized Euclidean distance, Minkowski distance, Mahalanobis distance, and the like. The smaller the distance between two vectors, the higher the degree of similarity between the two vectors.

  As a method for generating the transmitter group, a shortest distance method, a minimum distance method, a median method, a centroid method, a group average method, a K-means method, a neural network method, a self-organizing map method, and the like can be considered.

  In FIG. 1, the position estimation reference generation unit 5 is based on the feature vector record generated by the feature vector conversion unit 3 and the reference generation parameter set by the position estimation reference generation condition setting unit 4. A position reference record serving as a reference for estimating the position is generated and output to the position estimation reference storage unit 6.

  The position estimation reference generation unit 5 is a transmitter group in which radio wave reception status data (feature vectors) in a feature vector record are similar to each other based on all the feature vector records generated by the feature vector conversion unit 3 Is generated. For each generated transmitter group, a vector representative of the transmitter group (representative feature vector) is calculated. The number of elements of the representative feature vector (representative radio wave reception status data) is the total number of receivers installed in the field. Thereafter, a location reference record is generated by attaching a location ID representing a specific location in the field to each representative feature vector.

  Details of the procedure for generating the transmitter group, the method of calculating the representative feature vector, and the method of assigning the location ID will be described later. The location ID may be determined by the user by looking at the representative feature vector, and the location ID to be labeled need not be limited to the location where the receiver was placed.

  In FIG. 1, the position estimation reference storage unit 6 records the position reference record generated by the position estimation reference generation unit 5 and outputs the position reference record to the position estimation unit 7 and the alternative generation unit 12.

  FIG. 5 shows an example of the contents of the position estimation reference storage unit 6.

  The position estimation reference storage unit 6 stores a plurality of position reference records, and each position reference record is composed of the representative radio wave reception status data (representative feature vector) of the transmitter group and the location ID as described above. For example, in the first line of the position reference record, the representative feature vectors of the transmitter group at the location P1 are TNC01L1, TNC01L2, TNC01L3, TNC01L4,. A plurality of representative feature vectors may exist for the same location ID.

  In FIG. 1, the position estimation unit 7 is recorded in the feature vector record generated by the feature vector conversion unit 3, the reference generation parameter set by the position estimation reference generation condition setting unit 4, and the position estimation reference storage unit 6. Based on the obtained position reference record, a moving object record in which the position of the moving object is estimated is generated and output to the estimated position determination unit 10.

  The position estimation unit 7 calculates the similarity between the radio wave reception status data (feature vector) in the feature vector record and the representative radio wave reception status data (representative feature vector) in all the position reference records. The location ID of the location reference record including high representative radio wave reception status data is estimated as the location of the moving object. The position estimation unit 7 adds the estimated place ID to the feature vector record to generate a moving object record. The above is performed for all feature vector records.

  FIG. 6 shows an example of the moving object record generated by the position estimation unit 7.

  The mobile record is composed of a record number, mobile ID, location ID, time, and radio wave reception status data.

  The mobile body ID, time, and radio wave reception status data are obtained from the mobile body ID, time, and radio wave reception status data (feature vector) in the feature vector record. For example, the first line in FIG. 6 shows the radio wave reception status data (TNL1001, TNL2001, TNL3001,...) In the feature vector record in the first line in FIG. 4 as the representative in the position reference record in the first line in FIG. As a result of determining that the radio wave reception status data (TNCI01L1, TNCI01L2, TNCI01L3, TNCI01L4,...) Has the highest similarity, the location P1 is added to the feature vector record. That is, it is estimated that the mobile object PID001 was at the location P1 at time Time001 (between time Time001 and the reference time).

  In the above description, when the same location ID or the same radio wave reception status data continues to appear in a plurality of rows, those recorded together may be used as the mobile record. In this case, the mobile object record includes a record number, a mobile object ID, a place ID, a time, a staying time (the number of records (number of rows) × unit time), and radio wave reception status data. When they are grouped under the same location ID, the radio wave reception status data takes an average value or an intermediate value.

  FIG. 7 shows an example in which the same radio wave reception status data and the same location ID appear continuously in a plurality of rows.

  FIG. 8 shows an example in which records having the same radio wave reception status data are combined into one record in FIG. For example, since the radio wave reception status data of the record number 001 and the record number 002 in FIG. 7 are the same, these records are collected into the record of the record number 001 in FIG. At this time, the stay time ST001 of record number 001 in FIG. 8 is the difference between Time003 (time of record number 003) and Time001 (time of record number 001) in FIG. That is, it is estimated that the mobile object with the record number 001 in FIG. 8 is that the mobile object PID001 was at the location P1 from time Time001 to ST001, and the radio wave reception status data at this time is (1, 0, 0 , 0, 0).

  FIG. 9 shows an example in which records having the same location ID are combined into one record in FIG. In this example, the radio wave reception status data of the collected records is an average value of the radio wave reception status data in each record before being collected. For example, since the location IDs of record number 001 to record number 004 in FIG. 7 are the same, these records are combined into a record with record number 001 in FIG. At this time, the stay time ST′001 of record number 001 in FIG. 9 is the difference between Time005 (time of record number 005) and Time001 (time of record number 001) in FIG. Further, the radio wave reception status data of record number 001 in FIG. 9 is the average value (1, 0.3, 0, 0, 0) of the radio wave reception status data of record numbers 001 to 004 in FIG. That is, it is estimated that the mobile object with the record number 001 in FIG. 9 is that the mobile object PID001 was at the location P1 from time Time001 to ST'001, and the radio wave reception status data at this time is (1, 0.3 , 0, 0, 0).

  In FIG. 1, the movement constraint data storage unit 8 stores a mobile body constraint table that represents conditions (movement constraint conditions) that the mobile body must satisfy when moving in the field.

  FIG. 10 shows an example of the contents (moving body restriction table) of the movement restriction data storage unit 8.

  The moving body constraint table is represented by a square matrix of the total number of places × the total number of places, for example. A portion where a certain place ID (row) and a certain place ID (column) intersect represents a movement constraint condition that must be satisfied when moving between two places. For example, as a movement constraint condition, the minimum value of the number of places that must pass when moving between two places is conceivable.

  Here, when the observation spots (locations) P1 to P5 are arranged in the field F as shown in FIG. 11, the moving object restriction table is described as shown in FIG. 12, for example. Since it is possible to move from place P1 to P2 without passing through other places, 0 is stored in these intersections. Since the movement from the place P1 to the place P3 must always pass through the place P2 or P4, 1 is stored in these intersections.

  As a movement constraint condition, in addition to the minimum number of places existing between two places, a movement time required for movement between two places may be used. In this case, a portion where a certain place ID (row) and a certain place ID (column) intersect represents the shortest movement time required to move between two places.

  For example, when the shortest movement time between each place is as shown in FIG. 13, the moving object restriction table is described as shown in FIG. Since the movement from the place P1 to the place P2 takes 8 seconds, “8” is stored in these intersections.

  In addition to geographical movement restriction conditions, if the observation spot visited by the moving body is restricted according to the occupation and role of the moving body, and the accessible place is a part of the field, the restriction is 2 A value expressed as a value (0: movable, 1: unmovable) may be used as a movement constraint.

  As described above, the movement constraint data storage unit 8 has a geographical restriction such as the number of places that must be passed when moving between two places as described above, and a minimum movement time required when moving between two places. Only one of the time constraints may be included as the movement constraint condition of the moving body, or both of them may be included. Moreover, depending on the case, you may hold | maintain a movement constraint condition for every moving body separately.

  In the geographical conditions of FIG. 11 or FIG. 13, if the movement between the two places is all possible in both directions, the moving object restriction table becomes the target matrix (see FIG. 12 or FIG. 14). The lower right triangle part in the system contract table may be omitted.

  In FIG. 1, the position determination reference setting unit 9 is a condition (the condition necessary for the estimated position determination unit 10 to determine whether or not the location ID stored in the mobile object record output from the position estimation unit 7 is correct). Estimated position determination parameters) are set based on, for example, user selection or direct user input.

  For example, when determining whether or not the location ID is correct based on how many times the movement satisfying the movement constraint condition stored in the movement constraint data storage unit 8 is performed, the position determination parameter determines that the location ID is correct. This is the number of movements necessary to be performed. For example, the first place, the second place, the third place,... Moves to the nth place, the places up to the (n-1) th have been determined, and it is determined whether or not the nth place is correct. Think about the situation. In this case, there are a movement constraint from the (n-1) th place to the nth place, a movement restriction from the (n-2) th place to the nth place, a movement restriction from the (n-3) th place to the nth place, and so on. However, when it is sufficient to satisfy at least one of these constraint conditions (for example, the constraint on movement from the (n−1) th place to the nth place), the estimated position determination parameter indicating the number of movements is “1”.

  The estimated position determination parameter may be set in common for all the place IDs, or may be set for each place ID. Moreover, it may be set commonly for all the moving bodies, or may be set for each moving body.

  In FIG. 1, the estimated position determining unit 10 includes a moving object record indicating the estimated position of the moving object generated by the position estimating unit 7, a moving object constraint table stored in the movement constraint data storage unit 8, and a position determination criterion. Based on the estimated position determination parameter set by the setting unit 9, it is determined whether or not the location ID in the moving object record is correct.

  If the location ID in the mobile record is correct, an action history record (see FIG. 15) is generated from the mobile record and output to the action history storage unit 13. On the other hand, if the location ID in the mobile record is not correct, the mobile record is output to the alternative generation unit 12 and the alternative generation unit 12 outputs an alternative (location ID) sequence (one or more sequences of location IDs). First, the sequence length 1) is obtained. In this state, when the next mobile record is input to the estimated position determination unit 10 and the location ID in the mobile record is incorrect (when the alternative sequence is determined to be incorrect), the alternative generation unit The alternative sequence (sequence length 2) is acquired from 12, and the previously acquired alternative sequence (sequence length 1) is updated. In this state, the next moving body record is input to the estimated position determination unit 10, and until the location ID in the moving body record is determined to be correct (until the latest alternative sequence is determined to be correct), the above is performed. repeat. The estimated position determination unit 10 may sequentially acquire an alternative with a sequence length of 1 from the alternative generation unit 12 instead of acquiring an alternative sequence that sequentially increases in sequence length. If the location ID in the mobile record is determined to be correct (if the latest alternative sequence is determined to be correct), the stored alternative sequence is accepted and the mobile record that has been determined to be incorrect so far Are generated (for the number of alternatives included in the alternative sequence), and output to the behavior history storage unit 13. Details of the determination method by the estimated position determination unit 10 will be described later.

  In FIG. 1, an alternative generation condition setting unit 11 sets conditions (alternative generation condition parameters) necessary when the alternative generation unit 12 generates an alternative based on, for example, selection by the user or direct input by the user. Set.

  As an alternative generation condition parameter, for example, the minimum of the degree of similarity between the radio wave reception status data of the position reference record including the generated (stretched) alternative and the radio wave reception status data of the target mobile record must be satisfied Similarity can be considered. In this way, by pruning alternatives using the minimum similarity, it is possible to prevent the number of alternatives generated (stretched) from becoming enormous. All generated alternatives may be treated as candidates without branching alternatives.

  In FIG. 1, the alternative generation unit 12 includes a reference generation parameter set by the position estimation reference generation condition setting unit 4, a position reference record stored in the position estimation reference generation unit 6, and a moving object constraint data storage unit 8. , The estimated position determination parameter set by the position determination reference setting unit 9, the mobile record determined by the estimated position determination unit 10 that the estimated position is incorrect, and the alternative generation condition Based on the alternative generation condition parameters set by the setting unit 11, one or more alternative sequences satisfying the mobile body constraint table, the estimated position determination parameter, and the alternative generation condition parameters are generated, and one of them is generated. Output to the position determination unit 10. If an alternative sequence has already been generated, the alternative generation unit 12 extends the recorded alternative sequence by one, and outputs one of the sequences to the position determination unit 10. The alternative generation unit 12 continues to hold all the generated alternative sequences until the estimated position determination unit 10 determines that the location ID of the moving object record is correct (until the alternative sequence is determined to be correct). .

  For each generated (stretched) alternative (location ID), the alternative generation unit 12 receives the radio wave reception status data in the position reference record corresponding to the alternative and the radio reception of the target mobile record Calculate similarity to situation data. An alternative sequence having the highest similarity is selected and output to the estimated position determination unit 10. All the generated alternative sequences are stored until it is determined that the location ID in the moving body record input to the estimated position determination unit 10 is correct (until it is determined that the alternative sequence is correct). Each time one mobile object record is input from the estimated position determination unit 10 to the alternative generation unit 12, the stored alternative sequence is updated (the sequence is extended by one). When the estimated position determination unit 10 determines that the location ID in the moving object record is correct (when the alternative sequence is determined to be correct), the alternative generation unit 12 erases all generated alternative sequences. .

  If there is no alternative (location ID) that satisfies the alternative generation condition parameter (for example, minimum similarity) set by the alternative generation condition setting unit 11, for example, a message is returned and the program is stopped. In the above method, only the highest similarity is selected as the alternative sequence, but the estimated position determination unit 10 does not determine that the location ID in the next mobile record is correct. (If the alternative sequence is not correct), the alternative sequence having the second highest total similarity is extracted as the next candidate and output to the estimated position determination unit 10. The location ID in the body record may be determined. That is, the alternative sequence selection and the determination of the location ID in the mobile record ID may be repeated several times. The generation of the alternative sequence will be described in more detail later.

  In FIG. 1, the action history storage unit 13 records the movement history of the moving object as an action history record.

  FIG. 15 is a diagram illustrating the contents of the action history storage unit 13 as an example.

  The action history storage unit 13 includes a plurality of action history records. Each action history record includes a mobile object ID, a time, and a place ID. For example, the action history record in the first line indicates that the mobile object PID001 is at the location P1 when Time001 is (from Time001 to the reference time).

  In the position estimation apparatus as described above, first, a procedure (Step 501 to Step 505) of processing performed by the position estimation reference generation unit 5 will be described in detail below. Here, the Euclidean distance is used for calculating the similarity between vectors, and the K-means method is used for the transmitter group classification method. In the K-means method, the number k of groups to be generated is set in advance by the user. The total number of feature vectors is m, and the number of elements of each feature vector is the number n of receivers installed in the field.

Step 501: Determine initial values l ₁ ... L _k of representative vectors of groups (C ₁ , C ₂ ,... C _x ,... C _n ).
The initial value may be automatically determined using a random number, or may be set by the user in advance.
l ₁ = (l ₁₁ , l ₁₂ ,…, l _1n ), l ₂ = (l ₂₁ , l ₂₂ ,…, l _2n )…, l _k = (l _k1 , l _k2 ,…, l _kn )
However, (l ₁ ≠ l ₂ ≠ l ₃ ≠… ≠ _k )
l jth element of _ij group i

最もユークリッド距離が小さいグループCiにデータを分類する。

Step502: Calculate the similarity between all input feature vectors (p ₁ , p ₂ , ..., p _x , ..., p _m ) and the representative vectors of each group, and classify them into groups with high similarity .
Input feature vector p _x = (p _x1 , p _x2 , ..., p _xy , ..., p _xn ) and each group representative vector (l ₁ , l ₂ , ..., l _i , ..., Euclidean distance (E _1, E ₂ and _{l k), ···, E i} , ···, calculated using the following equation E _k).

Classify the data into the group Ci with the smallest Euclidean distance.

Step 503: The average of the radio wave reception status data classified in each group is obtained by the following formula, and the value is newly set as the representative vector of the group.

少なくともいずれかのグループについて上記式が成り立たない場合は、l_i=l'_i (l₁,l₂,・・・,l_i,・・・,l_k)として、Step502へ戻り、Step502からStep504まで繰り返す。 Step 504: It is determined whether or not the following expressions hold for all groups (C ₁ , C ₂ ,... C _x ,... C _n ).

If the above equation does not hold for at least one of the groups, return to Step 502 as l _i = l ′ _i (l ₁ , l ₂ ,..., L _i ,..., L _k ), and from Step 502 to Step 504 Repeat until.

Step 505: Representative vectors (l ′ ₁ , l ′ ₂ ,..., L ′ _i ,... Of each generated group (C ₁ , C ₂ ,... C _x ,... C _n ) Use l ' _k ) to label each group with a location. More details are as follows. Each element of the input feature vector p _x = (p _x1 , p _x2 ,..., P _xy ,..., P _xn ) is represented by each receiver (L ₁ , L ₂ ,..., L _y ,... L _n ) represents the number of detected transmitters.

Therefore, the element corresponding to L ₁ in the input feature vector includes the maximum number of transmitters (the number of transmitters held by the target user) r, and the other unit is 0. The location unit vector e ₁ = (r, Let 0,... 0,... 0) be a vector representing a location P ₁ near the center of the receiver L ₁ .

Location near the center of the receiver (L ₁ , L ₂ ,..., L _y ,... L _n ) (for example, the detection area of the receiver that does not overlap the detection range between the receivers) (P ₁ , P ₂ ,..., P _y ,... P _n ) and a representative vector (l) generated by Setp 504 (e ₁ , e ₂ ,..., E _x ,..., E _n ). The degree of similarity with ' ₁ , l' ₂ , ..., l ' _i , ..., l' _k ) is _obtained by the following equation.

Label where P _x Euclidean distance E _x is shown in the lowest (highest similarity) where the unit vector e _x the representative vector.

  Here, for a representative vector that does not satisfy the similarity set in advance by the user, a new observation spot other than the installation location of the receiver may be automatically generated and a new location label may be attached. In addition, the number of representative vectors labeled as locations near each receiver (this number may be determined for each location or may be determined uniformly) is set in advance by the user, and the number of representative vectors exceeding that number The vector may be automatically labeled at a new location or the next highest similarity location. In addition, a new location unit vector that considers not only the location near the center of the receiver but also the combination of all the receivers (for example, a location unit vector for a region where detection regions overlap between the receivers) is generated, and similar to it. The degree may be calculated. In addition, the user may correct the place ID automatically labeled.

  Next, a processing procedure performed by the estimated position determination unit 10 will be described with reference to FIG.

  FIG. 16 is a flowchart illustrating a processing procedure performed by the estimated position determination unit 10.

  Here, if the number of movements satisfying the movement constraint condition stored in the movement constraint data storage unit 8 is performed once, it is determined that the location ID in the target mobile record is correct. Further, a unit record is assumed to be 1 (s) using a mobile unit record including a record number, a mobile unit ID, a place ID, a time, and radio wave reception status data as shown in FIG. The estimated position determination is performed by sequentially reading the moving object records sorted in time series for each moving object. Record number i is the order of mobile records arranged in time series. The record number at the start of mobile observation is 1.

  In Step 1001, the mobile object record at the mobile object observation start time is read. The subsequent processing procedure differs depending on whether the read mobile record is a mobile record at the start of observation or a mobile record other than that. Therefore, in Step 1002, the process is branched into two patterns. In step 1004, it is determined whether or not to end the estimation position determination.

  If the read mobile record is the one at the start of observation (YES in Step 1002), the process proceeds to Step 1003. If the read mobile record is other than the observation start (NO in Step 1002) and the record number i is equal to or smaller than the total number of mobile record (YES in Step 1004), the process proceeds to Step 1005. If the record number i is larger than the total number of records (NO in Step 1004), the process is terminated.

In Step 1003, the location ID (E _{i = 1} ) of the mobile record when _{i = 1} is added to the sequence S with S _{i = 1} . S is a sequence for storing the place where the same moving object stayed in time series,
S = <S _{i = 1} , S _{i = 2} ,... S _{i = in} > (in: total number of records). Increment the value of i by 1 and return to Step 1002.

In Step 1005, the location S _{i-1 at} time i-1 stored in the sequence S and the location E _i of the moving object record at the record number i satisfy the movement constraint condition stored in the movement constraint data storage unit 8. (As described above, in this example, the number of movements that satisfy the movement constraint condition may be performed once). If the movement constraint condition is satisfied (YES in Step 1005), the process proceeds to Step 1006. If not satisfied (NO in Step 1005), the process proceeds to Step 1010.

In Step 1006, it is determined whether or not the alternative sequence A exists in the alternative generator 12. The alternative sequence A is selected from a plurality of alternative sequences generated by the alternative generation unit 12, and the place ID is stored in time series. Alternative sequence A represents the _{A = <A 1 ,A 2 ,···,A k} > (k: sequence length alternatives).

  If the alternative sequence A exists (YES in Step 1006), the process proceeds to Step 1009. If not (NO in Step 1006), the process proceeds to Step 1007.

In Step1007, record the mobile ID and the time (record number i-1 times of the mobile records) and the location IDS _i-1 as an action history record in the action history recording unit 13. When the mobile object record shown in FIG. 7 or FIG. 8 is targeted, in addition to these, the stay time (the stay time recorded in the record number i-1 of the mobile object record) is recorded. After recording, go to Step 1008.

In Step 1008, the location ID (E _i ) of the mobile record at record number i is added to sequence S as S _i . Increment the value of i by 1 and return to Step 1002.

  In Step 1009, all the place IDs included in the alternative sequence A are recorded in the action history recording unit 13 as an action history record so as to be in time series with the mobile body ID and the time. After recording, go to Step 1008.

  In Step 1010, the alternative generation unit 12 generates an alternative sequence in which the sequence length k (initially k = 0) is incremented by 1, and extracts a candidate alternative sequence A from the generated alternative sequence. The extracted alternative sequence A is passed to the estimated position determination unit 10. After this, the process proceeds to Step 1011. If the alternative sequence A cannot be extracted in Step 1010, the processing is stopped.

In Step 1011, it added to the sequence S k-th element A _k included in alternative sequence A _(elements of the re-tail alternative sequence) as S _i. If there is something in the sequence S that is different from the elements of the extracted alternative sequence A (if elements other than the rear end of the alternative sequence A are different from the previous alternative sequence A), the alternative sequence A The element in the sequence S is updated by the included element. Thereafter, the value of i is incremented by 1, and the process returns to Step 1002.

  Next, a processing procedure performed by the alternative plan generator 12 will be described.

  First, prior to describing the processing procedure by the alternative generation unit 12, the alternative sequence will be briefly described.

  In this embodiment, the alternative solution sequence is expressed by a tree structure. An example of the tree structure is shown in FIG. The tree structure is a data structure that starts from one node (root node) and repeats branching to deepen the hierarchy. A node generated by branching is called a child node, and a node from which the branching is made is called a parent node. A node can have only one parent node. For example, (T21, LT21) is a child node of (T11, LT11), and (T11, LT11) is a parent node of (T21, LT21). Each node stores a location ID (Tx: x is an arbitrary integer) and a similarity (LTy: y is an arbitrary integer). The stored similarity is (similarity stored in the parent node) + (radio wave reception status data in the position reference record having the location ID included in the child node, and radio wave reception status data in the target mobile record) The degree of similarity).

  In the alternative sequence, place IDs included in a child node at the lowest layer and an upper node of the child node excluding the root are arranged in the order of shallowness. For example, (T11, T21), (T11, T22), (T12, T23), (T13, T24), and (T13, T25) are alternative sequence included in the alternative tree of FIG. The alternative tree generated for the alternative sequence length k has a depth of k + 1, and the number of alternative sequences is the number of nodes of depth k + 1.

  Hereinafter, the processing procedure performed by the alternative generation unit 12 will be described in detail with reference to Step 1201 to Step 1206.

In Setp 1201, the latest location T ₀ (location ID last recorded in Step 1007 in FIG. 16) determined to be correct by the estimated position determination unit 10 is set as the parent node of the tree structure. At this time, the sequence length is set to k = 1 (in the state of k = 0, k is incremented by 1 so that k = 1).

In Step 1202, using the moving body constraint table stored in the movement constraint data recording unit 8, a location set T _k that satisfies the migration constraint condition is extracted for the location T ₀ , and each of the extracted locations is a child of the parent node. Store in the node.

The set T _k is expressed as T _k = (T _k1 , T _k2 ,..., T _kn ) (n: the total number of places satisfying the movement constraint condition for T ₀ ).

For example, when T ₀ = P ₁ , T ₁ = (P 1, P 2, P 4), where T ₁ is a set of places that can be directly moved from P 1 in the moving body constraint table of FIG.

Ｒ:移動体レコードの電波受信状況データ、mt_ki:場所集合Ｔ_ｋの要素iに対応する位置基準レコードにおける電波受信状況データ、l_k:T_kの親ノードに記録されている類似度（Ｔ_０＝0とする） Setp1203 calculates and saves the similarity between the radio wave reception status data in the position reference record corresponding to the location stored in each child node and the radio wave reception status data in the target mobile record using the following formula: To do.

R: radio wave reception status data of mobile record, mt _ki : radio wave reception status data in position reference record corresponding to element i of place set T _k , l _k : similarity recorded in parent node of T _k (T ₀ = 0)

For example, when the radio wave reception status data of the target mobile record is P = (0, 0, 0.8, 0.7, 0), the mobile record and each of T ₁ = (P1, P2, P4) The similarity with the element is calculated as follows. However, in the position estimation reference storage unit 6, as position reference records, (P1, 1, 0, 0, 0, 0), (P2, 0, 1, 0, 0, 0), (P4, 0, 0, Assume that 0, 1, 0) is stored.
l _p1p1 = 0 + (1 + 0 + 0.64 + 0.49 + 0) = 2.13
l _p1p2 = 0 + (0 + 1 + 0.64 + 0.49 + 0) = 2.13
l _p1p4 = 0 + (0 + 0 + 0.64 + 0.09 + 0) = 0.73

  In Step 1204, the similarity calculated in Step 1203 is normalized using the sequence length k (for example, the similarity is divided by the sequence length k). A child node that does not satisfy the minimum similarity set by the alternative generation condition setting unit 11 is deleted. Further, when a parent node having no child node is generated by deleting the child node, the parent node is also deleted.

  In Step 1205, the sequence including the place ID in the child node having the highest similarity among the results calculated in Step 1203 is output as the alternative sequence A to the estimated position determination unit 10. For example, in the above case, A = <P4>.

  In Step 1206, when the estimated position determination unit 10 determines that the alternative sequence A is correct (when it is determined that the location ID in the next mobile record is correct), the alternative tree is deleted. If it is not determined to be correct (see NO in Step 1005 in FIG. 16), k is incremented by 1, and the process returns to Step 1202.

  In the present embodiment described above, the position reference record generated by the position estimation reference generation unit 5 may be generated while sequentially acquiring the receiver acquisition data by the receiver, or the receiver acquisition acquired in advance. You may produce | generate based on data (sample data). In the former case, the position can be estimated in a state almost in real time.

  Moreover, in the estimated position determination part 10, when the estimation result by the position estimation part 7 is corrected, you may feed back a correction result to a position estimation reference | standard database using arbitrary methods. Thereby, the precision of the position estimation part 7 can be improved.

  Also, by generating a larger number of groups than the number of observation spots (number of receivers) set in advance and setting a portion where the reception ranges overlap as new observation spots, it is possible to generate many observation spots with a small number of receivers. Thereby, a field can be covered with few receivers.

  According to the present embodiment described above, even when a transmitter is received by a plurality of receivers, the position of the mobile body that seems to be the most appropriate in consideration of at least one of geographical constraints and time constraints. Can be easily estimated with less time and effort than conventional techniques.

It is a block diagram which shows the structure of the position estimation apparatus by embodiment of this invention. It is an example which shows the detection record of the transmitter by a receiver. It is a figure which shows the transmitter example which each mobile body hold | maintains. It is a figure which shows an example of a feature vector record. It is a figure which shows an example of a position reference record. It is a figure which shows an example of a mobile body record. It is a figure which shows the example which the same electromagnetic wave reception condition data and the same place ID appear continuously in multiple lines. In FIG. 7, it is a figure which shows the example which put together the record which has the same electromagnetic wave reception status data into one record. FIG. 8 is a diagram illustrating an example in which records having the same location ID are continuously combined into one record in FIG. 7. It is a figure which shows an example of a mobile body constraint table. It is a figure which shows the example of a field map. It is the figure which showed the geographical restrictions in the field map of FIG. 11 by the determinant. It is a figure which shows the example of a field map. It is the figure which showed the time restrictions in the field map of FIG. 13 with the determinant. It is a figure which shows an example of an action history record. It is a flowchart which shows the process sequence of an estimated position determination part. It is a figure which shows an example of the alternative produced | generated in an alternative production | generation part.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Receiver acquisition data memory | storage part 2 Mobile body possessed transmitter ID memory | storage part 3 Feature vector conversion part 4 Location estimation reference | standard generation condition setting part 5 Position estimation reference | standard generation part 6 Position estimation reference | standard storage part 7 Position estimation part 8 Movement restriction data storage Unit 9 position determination reference setting unit 10 estimated position determination unit 11 alternative generation condition setting unit 12 alternative generation unit 13 action history storage unit

Claims (14)

A position estimation device for estimating the position of a moving body having a transmitter in a field where a plurality of receivers are arranged:
A group of feature vectors representing the detection status of the transmitter by each of the plurality of receivers in unit time from detection record data in which the detection status of the transmitter by each of the receivers acquired in advance is recorded in time series. A feature vector generation unit to generate;
A representative feature vector generation unit that groups the group of feature vectors based on a predetermined grouping criterion, and generates a representative feature vector that represents the group for each group;
A position estimation reference generator for labeling a location in the field for each representative feature vector;
A position estimation device comprising:   The position estimation reference generation unit calculates a similarity between the representative feature vector and a place vector set for each of a plurality of locations in the field, and based on the similarity, The position estimation apparatus according to claim 1, wherein a location in the field is labeled.   The position estimation apparatus according to claim 1, further comprising a recording unit that records the detection status of the transmitter by each receiver in time series.   For each feature vector, a position estimation unit that calculates a similarity between the feature vector and each representative feature vector, and labels a location in the field with respect to the feature vector based on the similarity The position estimation apparatus according to claim 1, further comprising a position estimation device.   An estimated position determination unit configured to estimate whether or not a location labeled on the feature vector is correct based on a movement constraint condition that defines a condition that the moving body should satisfy in order to move between the locations in the field. The position estimation apparatus according to claim 4, further comprising:   6. The position estimation apparatus according to claim 5, wherein the movement constraint condition includes a number of places through which the moving body passes in order to move between places in the field.   The position estimation apparatus according to claim 5, wherein the movement constraint condition includes a movement time required for the moving body to move between locations in the field.   The position estimation apparatus according to claim 5, wherein the movement restriction condition includes a place in the field that is restricted by an attribute of the moving object.   An alternative generation unit that generates an alternative that satisfies the movement constraint condition when the estimated position determination unit determines that the location labeled on the feature vector is not correct. Item 9. The position estimation device according to any one of Items 5 to 8.   When there are a plurality of alternatives satisfying the movement constraint condition, the alternative generation unit detects representative feature vectors each labeled with a location according to each of the alternatives, and the labeled location is incorrect. Calculating the similarity between the determined feature vector and each detected representative feature vector, and selecting a place labeled with the representative feature vector having the highest similarity as the alternative. The position estimation apparatus according to claim 9. The alternative generation unit generates a sequence of alternatives with a sequence length increased by 1 each time the estimated position determination unit determines that the location labeled on the feature vector is incorrect.
The estimated position determination unit determines whether the location labeled on the feature vector is correct based on the movement constraint condition and the sequence of the alternative, and determines that the location labeled on the feature vector is correct. 11. The position estimation apparatus according to claim 9 or 10, wherein the position estimation apparatus accepts the alternative sequence.   The location determined to be correct by the estimated position determination unit or the sequence of the alternative received by the estimated position determination unit, and each of the alternatives included in the feature vector corresponding to the location or the sequence of the alternative The position estimation apparatus according to claim 11, further comprising an action history generation unit that generates an action history in the field of the mobile body from a feature vector corresponding to. A position estimation method for estimating the position of a mobile body having a transmitter in a field where a plurality of receivers are arranged:
Generating a group of feature vectors representing the detection status of the transmitter by each of the plurality of receivers in unit time from detection record data in which the detection status of the transmitter by each of the receivers acquired in advance is recorded in time series And
Grouping the group of feature vectors based on a predetermined grouping criterion and generating a representative feature vector representing the group for each of the groups;
For each representative feature vector, label a location in the field;
Position estimation method. A position estimation program for causing a computer to function as a means for estimating a position of a mobile object having a transmitter in a field where a plurality of receivers are arranged:
A group of feature vectors representing the detection status of the transmitter by each of the plurality of receivers in unit time from detection record data in which the detection status of the transmitter by each of the receivers acquired in advance is recorded in time series. Means for generating;
Means for grouping the group of feature vectors based on a predetermined grouping criterion and generating a representative feature vector representative of the group for each of the groups;
Means for labeling a location in the field for each representative feature vector;
A position estimation program for causing the computer to function as:

Images